Title : NSF9117 - Bio-Control by Neural Networks Type : Report NSF Org: BIO Date : September 1, 1991 File : nsf9117 ****************************************************************************** This File has been updated 10/31/96 to reflect the proper address of the: National Science Foundation 4201 Wilson Boulevard Arlington, VA 22230 For more information call: (703)306-1234 ****************************************************************************** BIO-CONTROL BY NEURAL NETWORKS Summary of a Workshop supported by the National Science Foundation George A. Bekey Computer Science Department University of Southern California and Peter G. Katona Program Director Bioengineering and Aiding the Disabled National Science Foundation Alexandria, Virginia May 16-18, 1990 Participating NSF Programs: Behavioral and Neural Sciences Bioengineering and Aiding the Disabled Engineering Systems Neuroengineering TABLE OF CONTENTS I. Introduction II. Workshop Agenda III. Summary of Presentations IV. Summary of Recommendations V. List of Attendees VI. References I. INTRODUCTION In the view of a number of investigators, there is an increasing dichotomy between engineering research in artificial neural networks and physiological research on neural control mechanisms. In order to determine the state of the art in both the biological and engineering view of bio-control by neural networks, to isolate the major difficulties that hinder communication and block progress in the field and to identify those areas where focused research might be most beneficial, NSF sponsored a small invitational workshop. The specific goals of the workshop were as follows: 1. To determine the state of the art in control of physiological systems by neural networks. How mature is this field? Can current models yield any insight into the structure and function of living control systems, or should they be viewed as input-output models, with little or no isomorphism to the nervous system? 2. To determine whether artificial neural networks, intended to mimic natural control systems, can be used to control systems that include biological components. Are we ready to design control systems that draw upon our knowledge of how natural systems behave? 3. To identify major difficulties that block progress in this field. Are the difficulties conceptual or experimental? Do we lack mathematical, computational, or experimental tools? Are there fundamental gaps in knowledge which hinder further application of artificial neural nets to living systems, either for model-building or for artificial control systems? The workshop was held on May 16-18, 1990 in Alexandria, Virginia. The 32 participants included six NSF program directors, two representatives from NIH, and 24 neural network researchers from both the biological and engineering communities. The conference was chaired by Dr. George Bekey, and sponsored by the Bioengineering and Aiding the Disabled Program. II. WORKSHOP AGENDA BIO-CONTROL BY NEURAL NETWORKS Wednesday, May 16, 1990 8:00 pm Social Gathering Thursday, May 17, 1990 8:30 am Introductions and Presentation of Workshop Goals George Bekey, University of Southern California, Conference Chairman NSF Program Directors: Peter Katona Kishan Baheti Duane Bruley Fred Heineken Nat Pitts Paul Werbos 9:00 am Control of Limb Movement James Houk, Northwestern University Emilio Bizzi, Massachusetts Institute of Technology 10:15 am Coffee Break 10:45 am Control of Cardiopulmonary Systems James Schwaber, DuPont Jack Feldman, University of California-Los Angeles John Lazzaro, California Institute of Technology Chi-Sang Poon, Massachusetts Institute of Technology 12:00 pm Lunch 1:30 pm Process Control by Neural Networks Lyle Ungar, University of Pennsylvania T. J. McAvoy, University of Maryland 2:30 pm Modeling Methodology Andrew Barto, University of Massachusetts 3:30 pm Coffee Break 4:00 pm Group Discussions I 6:30 pm Dinner 7:30 pm Control of Locomotion in a Simulated Insect Using Heterogeneous Neural Nets Hillel Chiel, Case Western Reserve University Friday, May 18, 1990 8:30 am Presentations from Groups; Open Discussions 10:00 am Coffee Break 10:30 am Control of Locomotion Sten Grillner, Karolinska Institute 11:15 am Methodology and Trends in Modeling Herb Rauch, Lockheed 12:00 pm Lunch 1:00 pm Group Discussions 2:30 pm Presentations from Groups; Summary of Recommendations 4:00 pm Adjourn III. SUMMARY OF PRESENTATIONS The Workshop featured a number of presentations which highlighted current research in various aspects of the field. The biological control issues discussed were primarily concerned with the neuromuscular and cardiopulmonary systems. The following paragraphs present brief summaries of the presentations. Further details on each topic can be obtained from the works listed in the References. [1-29] Emilio Bizzi (Brain and Cognitive Sciences Department, MIT) The Organization of Limb Motor Space in the Spinal Cord Results of new experiments were presented, relating to spinal cord mechanisms in the control of posture and movement. Microstimulation of the upper and middle layers of the frog's spinal cord (while the leg is placed in different positions) generated a force field with an equilibrium point. The implications of this field on the organization of the spinal cord were discussed. [1] [2] James Houk (Physiology Department, Northwestern University) Movement Control Based on an Adjustable Pattern Generation Model of the Cerebellum New results in certain aspects of motor control by the spinal cord and the cerebellum were discussed. Dr. Houk presented a novel view of Purkinje cells as bistable elements and cellular mechanisms for adjusting synaptic weights. Houk and Andrew Barto (Computer Science Department, Univ. of Massachusetts) are using these ideas for the design of a new model of cerebellar function. [3] Issues involving the neural control of locomotion were also discussed by Hillel Chiel and Sten Grillner. Hillel Chiel (Biology Department, Case Western Reserve University) Control of a Simulated Insect Dr. Chiel presented work (co-authored with Randall Beer, CWRU Department of Computer Engineering and Science) on an artificial neural network used for controlling the locomotion of a simulated insect. The network contained elements which were based on experimental observations, and were thus more complex than those utilized in most artificial neural network modeling. For example, some of the model neurons showed rhythmic bursts of activity ("pacemaker neurons") which were modulated by input from other model neurons. In addition, the architecture of the neural net controlling locomotion was based on studies of locomotion in insects done by Pearson and his colleagues. [4] The resulting network was capable of generating patterns of leg movements (gaits) which were statically stable. In addition, as the activation of a single command neuron was increased, the network generated a continuous range of faster, statically stable gaits. Furthermore, the network was robust in response to a wide variety of perturbations. [5] Additional neural networks were layered upon this locomotion controller, allowing the simulated insect to wander, follow the edges of obstacles to get around them, and locate and consume "food". [6] Thus, a network containing 78 model neurons and 158 synaptic connections was capable of exhibiting surprisingly complex behavior patterns. [7] Sten Grillner (Nobel Institute for Neurophysiology, Stockholm) Locomotion Control in the Swimming Eel The locomotor control system has a similar organization in all classes of vertebrates, with a spinal pattern generator network, which is controlled from the brainstem, and receive a powerful modulation from sensory afferents. In a simple vertebrate, the lamprey, with comparatively few neurones, the neuronal network has been detailed in terms of connectivity, cell properties, transmitters etc. This network has been simulated in a comparatively detailed fashion. It was found that without simulation, it was not possible to evaluate if the experimentally established network could account for the known locomotor behavior in terms of segmental and intersegmental coordination. [8] [9] [10] The autonomic nervous system was discussed by a number of researchers. James Schwaber (DuPont Neural Computation Group) Modeling the Baroceptor Reflex Dr. Schwaber presented an overview of cardio-respiratory control with an emphasis on modeling the control of blood pressure via the baroceptor vagal reflex. A fascinating aspect of this work was that in addition to the analytical and computational aspects of simulation of the neural networks involved in this system, presented by Wade Rogers (DuPont Neural Computation Group), the vagal baroceptor reflex has also been modeled in VLSI by John Lazzaro (Department of Electrical Computer Engineering, University of Colorado- Boulder). Dr. Rogers emphasized the complexity of the neurons involved in the reflex. The chip designed and fabricated by CalTech simulates the fibers in the vessel wall which give rise to the reflex. [11] [12] The interaction between biological experimentation, computer simulation, VLSI implementation, and analysis elicited a spirited discussion. Jack Feldman (Kinesiology Department, UCLA) Properties of Neurons in the Control of Respiratory Rhythms Dr. Feldman then discussed certain aspects of the control of respiration, primarily the generation of respiratory rhythms and the importance of various properties of the neurons involved in these systems. Distributed networks of coupled oscillators (i.e., the topology or spatial organization of the controlling neurons) were indicated as providing much of the robustness of the system. Dr. Feldman also discussed the temporal organization of the system and indicated that at least 12 different chemicals are associated with the neurotransmitters, and wondered whether they may control respiration in different time scales. The variety of neural action led him to urge engineers to use a larger variety of simulated neurons in their simulated networks. [13] Chi-Sang Poon (Biomedical Engineering Center, MIT) Multiple Optimization Criteria in Models of Respiratory Control Dr. Poon presented a mathematical model of the respiratory control system in which the input-output relationship of the brainstem respiratory controller was governed by an optimality criterion. The latter measured both deviation from steady state values of arterial PCO2 and the mechanical work output by the respiratory muscles. This model was shown to replicate many interesting characteristics of the respiratory control system -including the isocapnic state of exercise hyperpnea - without the need for an explicit exercise stimulus in the afferent path. [14] He then described a hybrid neural- computer experiment in which the respiratory optimization function was emulated by the visuo-motor neural circuits of the cerebral cortex, which served as a "proxy" of the brainstem neural network. [15] The results suggested that such compound optimization behavior was quite feasible within the CNS, both at the level of the brain stem and higher brain centers. He stressed that some adaptive neural network structures such as the Hopfield net may be relevant to the solution of this kind of biological optimization problems. [16] Lyle Ungar (Chemical Engineering Department, Univ. of Pennsylvania) and Thomas McAvoy (Chemical Engineering Department, Univ. of Maryland) Applications of Artificial Neural Nets in Chemical Process Control Various issues in the application of artificial neural nets in both feedforward and feedback control, inverse model adaptive control and other control algorithms were discussed. [17] [18] [19] Andrew Barto (Computer Science Department, Univ. of Mass.) On Computer Science Issues in Modeling Control Systems by Neural Nets Dr. Barto, who has been active in a number of interdisciplinary efforts on application of neural nets to engineering control problems, provided a general perspective on connectionist learning for control. Of particular interest was his view of the development of system models, indicating the appropriate role of engineering models, connectionist models and symbolic, AI-based models of physical systems. [20] Herbert Rauch (Palo Alto Research Lab, Lockheed) Comments on Modeling of Neural Networks Dr. Rauch, the Editor of the IEEE Transactions on Neural Networks, closed the formal presentations by presenting his views on some of the important research issues in the field of modeling of neural networks. These included questions on: (1) convergence properties of networks, (2) heuristic architectures for specific tasks, (3) adaptive architectures (self-organizing neural networks), and (4) new approaches, combining neural networks with such other techniques as expert systems or fuzzy logic for synergistic effects. George Bekey (Computer Science Dept, University of Southern California) made a brief presentation on neural network models (developed in collaboration with Thea Iberall) used for control of prehension in both human and robot hands. [21] [22] IV. RECOMMENDATIONS Much of the work of the workshop was accomplished in three subgroups which met following the major presentations. The groups first discussed the need for new biological data in engineering models of neural networks, as well as the need for engineering methodology in biological research. The discussions culminated in a series of recommendations for a research agenda, which is intended to support a possible new NSF initiative in this field. The recommendations are grouped into three categories: 1. General recommendations a) Cross-fertilization between disciplines is essential. There should be support for teams of researchers working in this field, in addition to single investigator support. b) Post-doctoral/sabbatical support could be used to place biologists in engineering labs and vice versa; perhaps these could be supported as supplements to existing projects. 2. Ways in which biology can assist control theory It was clear from many of the discussions that some engineering models of biological control systems are over-simplified, that they do not reflect much of the complexity of living organisms and, in particular, that the neuron models are primitive when compared with biological neurons. As a result, the following recommendations were made: a) More effort is needed in applying biologically inspired control strategies to non-biological applications: The control capabilities of "bio-controllers" need to be further examined. Mathematical developments for special cases are needed. b) More complex neuron models and architectures are needed for artificial neural networks: Model neurons should capture more of the richness of behavior patterns seen in biological experiments than the simple weighted-summer-with-sigmoid-nonlinearity that is most commonly used in artificial neural networks. Examples of neuronal properties to be considered: hysteresis, multi-modal behavior of single neurons in response to inputs, pre-synaptic inhibition and facilitation, and state-dependent effects. Neural network models should be heterogeneous, containing a variety of simulated neurons with different properties. More complex neural structures are needed to account for emergent behavior patterns as those found in living systems (e.g.: sensory-motor interactions, plant-controller interactions, distributed control paradigms). c) Improved methods for identifying, representing and analyzing hierarchical systems are needed: Identification of intrinsic hierarchical structures. Operation on multiple time scales. Minimization of information exchange between levels of hierarchy. Evolution of hierarchical systems. Theoretical bases for understanding the emergence of collective behavior form individual units and levels. d) Development and evaluation of new engineering adaptive control systems based on biological prototypes should be pursued: Enhancing living systems, e.g., prosthetics. Chemical process control, control of bioreactors. 3. Ways in which engineering can assist biology The discussions also revealed that many biologists in the Workshop are not using the best available tools from system theory in the formulation of models of biological control. In addition, significant gaps have been identified in existing methodologies that deal with complex biological systems. Hence, the following recommendations were made: a) There is a need for better instrumentation: Better sensors, especially parallel recording electrodes. Muscle-type actuators. Better motion monitoring equipment; tendon and contact force gauge implants and joint-angle monitoring implants. b) Better methods are needed for handling and abstracting voluminous data sets: Coding. Extracting information. Finding temporal structures in data. c) Adoption and improvements of systems methodologies are needed: System concepts; sensitivity and stability analysis; invertibility, decomposition. Methods of forming abstractions (moving from data to hypotheses), simplifications. Modeling techniques, classes of models, choosing models; modeling nonlinear systems. Computational and simulation tools, including object- oriented methodologies. Methods of real time planning and control. Methods of representing and analyzing hysteresis. System level hypotheses to direct experiments. V. LIST OF ATTENDEES Dr. Panos J. Antsaklis Department of Electrical and Computer Engineering University of Notre Dame Notre Dame, IN 46556 (219) 239-5792 (219) 239-8007 (FAX) Dr. Kishan Baheti Control Theory National Science Foundation Room 1151, ECS/ENG 1800 G Street, N.W. Washington, DC 20550 (202) 357-9859 Dr. Andrew G. Barto Department of Computer and Information Science University of Massachusetts Amherst, MA 01003 (413) 545-2109 (413) 545-1249 (FAX) Dr. George A. Bekey Computer Science Department University of Southern California Los Angeles, CA 90089 (213) 740-4501 (213) 740-7285 (FAX) Dr. Emilio Bizzi Department of Brain & Cognitive Sciences E25-526 Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge, MA 02139 (617) 253-5759 (617) 258-5342 (FAX) Dr. Eugene Bruce Department of Biomedical Engineering Case Western Reserve University Cleveland, OH 44106 (216) 368-6590 (216) 368-4969 (FAX) Dr. Duane Bruley (Biochemical Engineering National Science Foundation) Current address: School of Engineering California Polytechnic Institute San Luis Obispo, CA 93407 (805) 756-2131 Dr. Daniel Bullock Center for Adaptive Systems Boston University 11 Cunnington Street Boston, MA 02215 (617) 353-9486 (617) 353-2053 (FAX) Dr. Hillel Chiel Department of Biology Case Western Reserve University 2080 Adelbert Road Cleveland, OH 44106 (216) 368-3846 (216) 368-4672 (FAX) Dr. Howard J. Chizeck Department of Systems Engineering Case Western Reserve University Crawford Hall, 6th Floor Cleveland, OH 44106 (216) 368-3836 Dr. Jack Feldman Systems Neurobiology Laboratory Department of Kinesiology, UCLA 405 Hilgard Avenue Los Angeles, CA 90024-1568 (213) 825-0954 (213) 825-6616 (FAX) Dr. Apostolos P. Georgopolous Department of Neuroscience Johns Hopkins University 725 N. Wolfe Street Baltimore, MD 21205 (301) 955-8334 (301) 955-3623 (FAX) Dr. Sten Grillner Karolinska Institute The Nobel Institute for Neurophysiology Box 60400, S-104 Stockholm, Sweden 011-46-8-336059 011-46-8-349544 (FAX) Dr. William Heetderks Room 916, Federal Building National Institute of Neurological Disorders and Stroke National Institutes of Health Bethesda, MD 20892 (301) 496-5745 (301) 402-1501 (FAX) Dr. Fred Heineken Biotechnology National Science Foundation Room 1132, BCS/ENG 1800 G Street, N.W. Washington, DC 20550 (202) 357-7955 (202) 397-9803 (FAX) Dr. James C. Houk Department of Physiology Ward Building 5-319 Northwestern University Medical School 303 E Chicago Avenue Chicago, IL 60611 (312) 503-8219 (312) 503-5101 (FAX) Dr. Peter Katona Bioengineering and Aiding the Disabled National Science Foundation Room 1132 1800 G Street, N.W. Washington, DC 20550 (202) 357-7955 (202) 397-9803 (FAX) Dr. John Lazzaro Department of Electrical and Computer Engineering University of Colorado at Boulder Campus Box 425 Boulder, Co 80309 (303) 492-2896 (303) 492-3674 (FAX) Dr. Lina Massone MIT Building E25-526 77 Massachusetts Avenue Cambridge, MA 02439 (617) 253-5769 (617) 253-8000 (FAX) Dr. Thomas McAvoy Department of Chemical Engineering University of Maryland College Park, MD 20742 (301) 454-2432 (301) 454-0855 (FAX) Dr. Richard Nakamura MCT Program National Institute for Mental Health Room 11-105 5600 Fishers Lane Rockville, MD 20857 (301) 443-3948 (301) 443-4822 (FAX) Dr. Cliff Parten University of Tennessee 615 McCallie Avenue Chattanooga, TN 37403 (615) 755-4386 or 622-4642 (615) 622-4625 (FAX) Dr. Nathan Pitts Neural Mechanisms of Behavior National Science Foundation Room 320, BNS/BBS 1800 G Street, N.W. Washington, DC 20550 (202) 357-7040 Dr. Chi-Sang Poon Biomedical Engineering Center Room 20A-126 Massachusetts Institute of Technology Cambridge, MA 02139 (617) 258-5405 (617) 253-2514 Dr. Herb Rauch Palo Alto Research Lab Lockheed 92-20/254E 3251 Hanover Street Palo Alto, CA 94304 (415) 424-2704 (415) 424-2662 (FAX) Dr. David A. Robinson Room 355 Woods Research Building The Wilmer Institute The Johns Hopkins Hospital 600 N Wolfe Street Baltimore, MD 21205 (301) 955-3587 (301) 955-0867 (FAX) Dr. Wade T. Rogers DuPont Neural Computation Group DuPont Experimental Station E-357 Wilmington, DE 19880-0357 (302) 695-7945 (302) 695-2747 (FAX) Dr. James Schwaber DuPont Neural Computation Group DuPont Experimental Station E-352 Wilmington, DE 19880-0352 (302) 695-7136 (302) 695-9631 (FAX) Dr. Robert J. Sclabassi Department of Neurosurgery University of Pittsburgh Pittsburgh, PA 15213 (412) 692-5093 (412) 692-5287 (FAX) Dr. Rajko Tomovic Faculty of Electrical Engineering Belgrade University Bulevar Revolucije 73 P.O. Box 816 1101 Belgrade, Yugoslavia 011-38-11-335-329 Dr. Lyle H. Ungar Chemical Engineering Department University of Pennsylvania 220 S 33rd Street Towne Building, 311-A Philadelphia, PA 19104-6393 (215) 898-7449 (215) 898-1130 (FAX) Dr. Paul Werbos Neural Networks National Science Foundation Room 1151, ECS/ENG 1800 G Street, N.W. Washington, DC 20550 (202) 357-9618 VI. REFERENCES 1. Massone, L., and Bizzi, E., "A neural network model for limb trajectory formation," Biological Cybernetics, 61:417-425, 1989. 2. Bizzi, E., Mussa-Ivaldi, F.A., and Giszter, S.F., "Motor space coding in the central nervous system," 55th Symposium on Quantitative Biology: The Brain, New York: Cold Spring Harbor Laboratory (CSH), 1990. 3. Houk, J.C., Singh, S.P., Fisher, C., and Barto, A.G., "An adaptive sensorimotor network inspired by the anatomy and physiology of the cerebellum," Neural Networks for Control, Chapter 15, W.T. Miller, R.S. Sutton and P. J Werbos, (Eds), Cambridge: MIT Press, 1990. 4. Beer, R.D., Chiel, H.J., and Sterling, L.S., "A biological perspective on autonomous agent design," Robotics and Autonomous Systems, 6:169, 1990. 5. Chiel, H.J., and Beer, R. D., "A lesion study of a heterogenous artificial neural network for hexapod locomotion," Proc. IJCNN, I:407-413, 1989. 6. Selverston, A.I., "A consideration of invertebrate central pattern generators as computational data bases," Neural Networks, 1:109-117, 1988. 7. Houk, J.C., "Cooperative control of limb movements by the motor cortex, brainstem and cerebellum," Models of Brain Function, pp. 309-325, RMJ Cotterill, ed., Cambridge, Great Britain: Cambridge University Press, 1989. 8. Grillner, S., "Control of locomotion in bipeds, tetrapods and fish," The Handbook of Physiology, Sec. 1, Vol. 2: The Nervous System, Motor Control, pp. 1179- 1236, V.B. Brooks, (Ed.), Maryland: Waverly Press, 1981. 9. Matsushima, T. and Grillner, S., "Intersegmental coordination of undulatory movements - a 'trailing oscillator' hypothesis," NeuroReport 1, pp. 97-100, 1990. 10. Grillner, S., Wallen, P., and Brodin, L., "Neuronal network generating locomotor behavior in lamprey: Circuitry, transmitters, membrane properties, and simulation," Ann. Rev. Neurosci., 14, pp. 169-99, 1991. 11. Advanced Research in VLSI International Conference 1991, Carlo Sequin, MIT Press, Chap: Silicon Ba receptors modeling cardiovascular pressure transduction in ANALOG VLSI, Lazarro, John, Schwaber, James and Rogers, Wade. 12. Schwaber, J.S., Paton, J.F., Spyer, K.M., and Rogers, W.T., "Dynamic model-based controller in the brain using temporal filtering and pattern generation," (unpublished manuscript). 13. Feldman, J.L., "Emergent properties of neural mechanisms controlling breathing in mammals," Respiratory Muscles and their Control, Alan Liss, Inc., 1987. 14. Poon, C.S., "Ventilatory control in hypercapnia and exercise: optimization hypothesis," Journal of Applied Physiology, 62:2447-2459, 1987. 15. Poon, C.S. and Younes, M., "Optimization model of ventilatory control: interactive computer simulation," FASEB J.2: A1287, and "Optimization behavior of brainstem respiratory neurons: a cerebral neural network model," (submitted). 16. Poon, C.S., "Adaptive neural network that subserves optimal homeostatic control of breathing," (submitted). 17. McAvoy, T.J., "Modeling chemical process systems via neural computation," IEEE Control Systems Magazine, Vol. 10, No. 3, pp. 24-29, April 1990. 18. McAvoy, T.J., "Use of neural nels for dynamic modeling and control of chemical process systems," Computers Chem. Engr., Vol. 14, No. 4/5, pp. 573-583, 1990. 19. Ungar, L.H., and Psichogios, D.C., "Direct and indirect model based control using artificial neural networks." 20. Barto, A., "Connectionist networks for control," Neural Networks for Control, T. Miller, R.S. Sutton, and P.J. Werbos (Eds), Cambridge, MIT Press, 1990. 21. Iberall, T., Liu, H., and Bekey, G.A., "Building a generic architecture for robot hand control," IEEE International Conference on Neural Networks, Vol. II, pp. 567-574, San Diego, CA, July 1988. 22. Yeung, D.Y., and Bekey, G.A., "Using a context-sensitive learning network for robot arm control," Proc. 1989 IEEE International Conference on Robotics and Automation, pp. 1441-1447, May 1989. 23. Bullock, D. and Grossberg, S., "Neural dynamics of planned arm movements: Emergent invariants and speed- accuracy properties during trajectory formation," Psychological Review, 95, pp. 49-90, 1988. 24. Bullock, D. and Grossberg, S., "Spinal network computations enable independent control of muscle length and joint compliance," Advanced neural computers, pp. 349-356, R. Eckmiller (ed.), Amsterdam: Elsevier, 1990. 25. Grossberg, S. and Kuperstein, M., Neural dynamics of adaptive sensory-motor control, Expanded Edition, New York: Pergamon Press, 1989. 26. Houk, J.C., "Outline for a theory of motor learning," NATO ASI on Tutorials in Motor Neuroscience, G.E. Stelmach (Ed), Kluwer Acad. Publishers, (In Press), 1991. 27. Kawato, M., Furakawa, K., and Suzuki, R., "A hierarchical neural-network model for control and learning of voluntary movement," Biological Cybernetics, 57, pp. 169- 185, 1987. 28. Massone, L. and Bizzi, E., "On the role of input representation in sensory motor mapping," Proc: IJCNN, 1: pp. 173-176, 1989. 29. Robertson, R.M. and Pearson, K.G., "Neural networks controlling locomotion in locusts," Model Neural Networks and Behavior, Chap. 2, pp. 21-35, A.L. Selverston (Ed), New York: Plenum Press, 1985. The opinions expressed at the workshop are not necessarily those of the National Science Foundation. The National Science Foundation has TDD (Telephonic Device for the Deaf) capability, which enables individuals with hearing impairment to communicate with the Division of Personnel and Management about NSF programs, employment, or general information. This number is (202) 357-7492. P.T. 42 K.W. 0710035 1004000